DRUG DISCOVERY AND RESISTANCE Pre-treatment mycobacterial sputum load influences individual on- treatment measurements Andreas H. Diacon a, b, * , Lize van der Merwe c, e , Anne-Marie Demers a , Florian Von Groote-Bidlingmaier b , Amour Venter c , Peter R. Donald d a Department of Biomedical Sciences, Faculty of Medicine and Health Sciences, Stellenbosch University, Cape Town, South Africa b Task Applied Science, Bellville, Cape Town, South Africa c MRC Centre for Molecular and Cellular Biology, Faculty of Medicine and Health Sciences, Stellenbosch University, Cape Town, South Africa d Department of Paediatrics and Child Health, Faculty of Health Sciences, Stellenbosch University, Cape Town, South Africa e Department of Statistics, Faculty of Natural Sciences, University of the Western Cape, Cape Town, South Africa article info Article history: Received 5 May 2014 Accepted 31 August 2014 Keywords: Tuberculosis Antituberculosis treatment Colony forming units Time to culture positivity summary Time to culture positivity (TTP) in liquid medium is now widely available as a measure of viable mycobacterial sputum load. TTP correlates well with and could replace colony-forming unit (CFU) counting in studies of antituberculosis drug effects. We investigated the influence of the pre-treatment mycobacterial sputum load on 4428 CFU measurements obtained within the first 14 days of treatment. Using a prediction model we show that pre-treatment CFU counts contribute 29% to the variation of on- treatment CFU counts and increase the precision of the prediction of on-treatment CFU from TTP by 12%. On the other hand, pre-treatment TTP contributed only 12% to the variation of on-treatment TTP and only added 2% to the prediction of TTP from CFU. We conclude pre-treatment measurements are covariates that can enhance the accuracy of statistical estimates of treatment effects, particularly when measured by CFU counts. © 2014 Elsevier Ltd. All rights reserved. 1. Introduction Colony forming unit (CFU) counts of Mycobacterium tuberculosis and time to positive (TTP) culture in liquid medium are regularly employed measurements of sputum mycobacterial load. Antitu- berculosis treatment effects can be estimated by comparing mea- surements obtained under treatment to measurements collected before treatment [1e3]. At present smears graded at least 1þ on the WHO/IUATLD scale are required for participation in studies of early antituberculosis treatment effects. This maximizes the chance for positive sputum cultures over the time needed to detect and quantify treatment activities. It is generally accepted that sputum contains variable pro- portions of mycobacterial subpopulations with a spectrum of metabolic activity, and that different drugs target specific sub- populations [4,5]. A predominant population of rapidly metabo- lizing, extracellular mycobacteria is believed to determine the magnitude of CFU and TTP before and during the early stages of treatment [6,7]. This creates a potentially significant but underap- preciated source of error in the measurement of treatment effects. The standard requirement of “sputum smear grade 1þ” reflects a more than 100-fold range from the lowest to the highest possible pre-treatment sputum mycobacterial burden in study subjects. An abundance of metabolically highly active mycobacteria in a subject, for instance, might favour any agent targeting this bacterial phenotype but underestimate the activity of agents that target other bacterial phenotypes. The pre-treatment mycobacterial load could thus be an independent predictor for drug activity measured by a fall in CFU or increase of TTP over time [8,9]. This problem has been largely ignored in antimycobacterial activity studies because culture-based measurements of load become available too late to be prospectively controlled for. At present it is uncertain whether the pre-treatment load influences quantitative estimates of drug activity. To investigate this question we constructed prediction models for CFU and explored whether pre-treatment CFU and pre-treatment TTP values alter the model's prediction of CFU counts obtained within the first 14 days of treatment, ignoring individual patient level factors and the treat- ment received in order to focus on what percentage of the variation * Corresponding author. Department of Biomedical Sciences, Faculty of Medicine and Health Sciences, Stellenbosch University, PO Box 19063, 7505 Tygerberg, South Africa. Tel.: þ27 21 917 1044; fax: þ27 21 938 9476. E-mail address: ahd@sun.ac.za (A.H. Diacon). Contents lists available at ScienceDirect Tuberculosis journal homepage: http://intl.elsevierhealth.com/journals/tube http://dx.doi.org/10.1016/j.tube.2014.08.015 1472-9792/© 2014 Elsevier Ltd. All rights reserved. Tuberculosis 94 (2014) 690e694